Extraction of decision rules from class borders in ordinal classification problems

The ordinal classification problem consists in dividing full set of alternatives by several classes. These classes are ordered according to the classification goal. The set of alternatives contains all the possible combinations of estimates by several criteria. The criteria scales are also ordered according to classification goal. Due to the problem orderliness the resulting classes could be defined by their borders: other alternatives belonging to the class would be in domination relation with some of the border alternatives, i.e. would lay between the borders. There is a hypothesis that we can extract decision rules by analyzing class borders. These rules are close to those implicitly used by expert while performing classification. This paper introduces a formal definition of a decision rule and proposes an algorithm to represent a class border as a set of formal decision rules.